Governance risks at OpenAI
A New Yorker‑sourced investigation summarized in coverage alleges problematic internal behavior and concentrated decision‑making at OpenAI, raising questions about governance at a firm many robotics teams rely on for models. (ynetnews.com) Those governance concerns feed directly into vendor‑dependency and trust debates for embodied‑AI projects.
The company that sells some of the most important artificial intelligence tools in the world is again facing a very old problem: who, exactly, can tell its chief executive no. A new New Yorker investigation says OpenAI’s 2023 board revolt grew out of private concerns about Sam Altman’s conduct and truthfulness, not just a sudden personality clash. (newyorker.com) That matters because OpenAI was built with an unusual legal design. OpenAI says its nonprofit foundation controls its for-profit business, which now operates as a public benefit corporation rather than a standard startup chasing shareholder value alone. (openai.com) That structure was supposed to act like a brake pedal. OpenAI’s 2019 explanation for creating OpenAI LP said investors and employees could earn only capped returns, while the nonprofit kept ultimate control over the mission. (openai.com) Then the brake pedal failed in public. On November 17, 2023, OpenAI’s board announced that Altman was out as chief executive because it no longer had confidence in his ability to lead the company. (openai.com) Four days later, the company was close to breaking apart instead. Axios reported that an overwhelming majority of OpenAI’s 700-plus employees signed a letter telling the board to resign and saying they could join Microsoft with Altman. (axios.com) By November 29, 2023, Altman was back. OpenAI announced his return with a new initial board, which showed that the formal board had the legal power to fire him, but not the practical power to keep him out once employees and Microsoft lined up behind him. (openai.com) The new reporting makes that episode look less like a one-week corporate soap opera and more like a governance stress test. If board members believed the chief executive was withholding information, and still could not hold the line, the real center of power may have been the person, not the structure. (newyorker.com) OpenAI has spent the two years since then tightening its official story. The company says its October 28, 2025 update left the nonprofit foundation in control while converting the operating business into OpenAI Group PBC, a public benefit corporation meant to balance mission and commercial scale. (openai.com) OpenAI also says that structure can support more than $100 billion in resources while keeping mission-focused governance at the center. That is a much bigger machine than the one that imploded in November 2023, which means any weakness in oversight now sits on top of far more money, compute, and customer dependence. (openai.com) For robotics teams, this is not abstract boardroom gossip. A robot that uses OpenAI models for planning, speech, or vision is tying part of its product roadmap to a vendor whose key decisions can still hinge on a small circle of executives and directors. (openai.com) That creates a simple business risk. If the supplier changes pricing, access rules, safety policies, or product priorities after an internal power struggle, the robot company does not just get a bad press cycle; it may have to rebuild core behavior on a new model stack. (openai.com) So the real question in this story is no longer whether OpenAI has a nonprofit on top of a for-profit. The question is whether any paper structure can reliably govern a company whose products are becoming infrastructure before its internal checks have proved they can survive a fight with its own chief executive. (newyorker.com)
